Add IR to cuda backend

[govindchari]

Added iterative refinement to cuda backend. Performance compared to main on the MM problems is below:

Branch	Commit	Solved	Total iterations
main no IR	9acc233	128 / 138	3,202
gc/add-ir-cuda-backend	9826cdd	131 / 138	3,031

IR branch solved 3 additional problems: LISWET1, LISWET7, QSHELL.

Performance on MPC problems is below:

Branch	Solved	Total iterations
main no IR	28 / 64	2,603
gc/add-ir-cuda-backend	43 / 64	2,605

IR solved 15 additional MPC problems and did not lose any solved by main.

Newly solved by IR:
aircraft_1, aircraft_11, aircraft_2, aircraft_4, ballOnPlate_1, ballOnPlate_2, ballOnPlate_3, binaryDistillationColumn_1, binaryDistillationColumn_2, pendulum_1, pendulum_2, pendulum_3, springMass_2, springMass_3, springMass_4.

[github-actions]

Download benchmark artifacts

Benchmark Summary

• Baseline solved: 136 problems
• Diff branch solved: 136 problems

Runtime regressions (> 5.0%)

• CONT-300: diff=22.6835s, baseline=21.2942s, Δ=+6.5%
• HS118: diff=0.0001s, baseline=0.0001s, Δ=+20.5%
• HS51: diff=0.0000s, baseline=0.0000s, Δ=+33.3%
• PRIMALC8: diff=0.0048s, baseline=0.0046s, Δ=+5.6%
• QFFFFF80: diff=0.1263s, baseline=0.1178s, Δ=+7.2%

Runtime improvements (> 5.0%)

• DPKLO1: diff=0.0001s, baseline=0.0002s, Δ=-16.3%
• GENHS28: diff=0.0000s, baseline=0.0000s, Δ=-16.7%
• TAME: diff=0.0000s, baseline=0.0000s, Δ=-10.0%